Increasingly intricate applications like artificial intelligence demand stronger and more energy-intensive computers. Optical computing is suggested as a way to enhance speed and energy efficiency; however, it remains unactualized due to various challenges. A novel design framework known as diffraction casting aims to mitigate these issues. It introduces fresh ideas into the realm of optical computing, potentially making it more attractive for future computing devices.
Increasingly intricate applications like artificial intelligence demand stronger and more energy-intensive computers. Optical computing is suggested as a way to enhance speed and energy efficiency; however, it remains unactualized due to various challenges. A novel design framework known as diffraction casting aims to mitigate these issues. It introduces fresh ideas into the realm of optical computing, potentially making it more attractive for future computing devices.
All modern computer devices, whether smartphones or laptops, rely on electronic technology. However, this technology comes with its downsides, particularly the substantial heat generated as performance increases, in addition to reaching the fundamental limits of current manufacturing methods. Consequently, researchers are investigating alternative computation methods to overcome these issues while possibly introducing new functionalities.
One potential solution is rooted in an idea that has been around for many years but hasn’t yet achieved commercial success: optical computing. This technology utilizes the speed of light and its ability to interact with various optical materials without causing heat. Furthermore, a wide range of light waves can simultaneously traverse materials without interference, theoretically enabling the creation of a highly parallel, rapid, and energy-efficient computer.
“In the 1980s, Japanese researchers investigated an optical computing technique known as shadow casting that could perform basic logical operations. However, their implementation relied on substantial geometric optical structures, which could be compared to the bulky vacuum tubes of early digital computers. They were conceptually functional but lacked the flexibility and compatibility needed for practical applications,” stated Associate Professor Ryoichi Horisaki from the Information Photonics Lab at the University of Tokyo. “We present an optical computing framework called diffraction casting, which builds on shadow casting. While shadow casting relies on the interaction of light rays with various geometries, diffraction casting uses the properties of light waves itself, leading to optical elements that are more spatially efficient and functionally versatile, essential for a universal computer. We conducted numerical simulations that produced promising results, using small 16-by-16 pixel black-and-white images as inputs, smaller than typical smartphone icons.”
Horisaki and his team put forward an all-optical system, which means only the final output is converted to electronic and digital format; all prior steps remain optical. Their concept starts with an image as a data source—suggesting its potential for image processing—while other data types, especially those used in machine learning, could also be represented graphically. They liken this process to layering in an image editing software like Adobe Photoshop: there’s an input layer (the source image) with additional layers that can modify or manipulate the underlying content. The output from the top layer is processed through the combined layers. In this approach, light passes through these layers, casting an image (hence the term “casting” in diffraction casting) onto a sensor, ultimately converting it into digital data for storage or user presentation.
“Diffraction casting is simply one component in a theoretical computer based on this principle. It might be more accurate to consider it as an extra part rather than a complete substitute for existing systems, similar to how graphical processing units are specialized components for graphics, gaming, and machine learning tasks,” noted lead author Ryosuke Mashiko. “I estimate it will take about another 10 years before it becomes commercially viable, as substantial work remains to be done on the physical realization, which, despite being grounded in solid research, has yet to be constructed. Currently, we can demonstrate the effectiveness of diffraction casting in executing the 16 fundamental logic operations crucial to much information processing, and there’s also the potential to extend our system into emerging areas like quantum computing. Time will reveal the results.”
